Expected goals (xG) in betting: what it shows and where it misleads

What xG measures

Every shot gets a number between 0 and 1: the share of similar shots in the past that ended up as a goal. Stats Perform, whose Opta data feeds many broadcasters and stats sites, describes it as the likelihood that a chance is scored, based on information about similar historical shots. If 1,000 shots from a certain position and situation produced 100 goals, a new shot like that is worth 0.10 xG.

What counts as "similar" depends on the model. Common inputs are:

• distance and angle to the goal;
• body part (foot or head);
• the type of pass before the shot (cross, through ball, cut-back);
• the situation: open play, counter-attack, set piece or penalty;
• in more detailed models, the position of the goalkeeper and defenders.

A penalty sits at around 0.76 xG in Stats Perform's model, a header from a cross at the edge of the box often below 0.05. A team's xG for a match is simply the sum of all its shots.

Where to find xG

You no longer need a paid data subscription to see xG. Many free football stats websites, league apps and broadcasters show xG per match, per team and per player, and some show a timeline of every shot. The data comes from a handful of providers, and each provider uses its own model.

That has a practical consequence: the same match can have different xG values on different sites. One model may give a team 1.6 xG, another 1.3, because they weight headers, deflections or defender positions differently. If you compare numbers, always compare them from the same source.

Why public xG is already in the odds

Here is the part most "xG betting" articles skip. A number that millions of people can see for free on match day is not a secret. Bookmakers set their prices with professional data teams, and the biggest bettors in the market work with richer data still: tracking data, line-ups, injuries, travel, weather and their own models. When sharp money moves a price, the bookmakers follow.

Research on football betting markets finds that closing odds are, on the whole, well calibrated: across many matches, teams priced at 50% win about half the time. That is what we explain in betting market efficiency. An xG table from last weekend is part of the information those prices already contain. Reading it does not give you an edge on its own; at best it helps you understand why the odds are where they are.

Smaller leagues, lower divisions and women's football are less heavily analysed, so prices there can be less sharp. But they also have lower limits, thinner data and bigger bookmaker margins (see bookmaker margin), so any gap is smaller than it looks.

Example: how xG gets misused

Team A has lost its last three matches 0–1, 1–2 and 0–1. On xG it "won" all three: 1.9 to 0.8, 2.2 to 1.1 and 1.7 to 0.9. A bettor reads this as "unlucky, due a win" and backs Team A at odds of 2.40 in the next match.

What the bettor needs is a probability higher than the one in the odds. Odds of 2.40 imply 1 ÷ 2.40 = 41.7% (more in implied probability). The expected value of a €10 bet is:

EV = p × 2.40 − 1

• If Team A's true chance is 41%, EV = 0.41 × 2.40 − 1 = −1.6%, or −€0.16 per €10.
• If it were 50%, EV would be +20%, or +€2.00 per €10.

The bettor's case rests on the second line. But three good xG games are visible to everyone, including the people who set the 2.40. If the market believed Team A was really a 50% shot, the price would be nearer 2.00. A +20% edge on a top-league match result is not something a free table hands you.

The common mistakes behind this example:

• Tiny samples. Three matches is a handful of shots. A single penalty adds about 0.76 xG and can turn a match's xG story around.
• Ignoring game state. A team that leads early often sits deep and lets the opponent shoot from distance. The trailing team piles up xG while the leader controls the game.
• Red cards and line-ups. xG from a match played with ten men, or with a rotated side, says little about the next one.
• Gambler's thinking. "Due a win" is the gambler's fallacy in football clothes. Our guide to cognitive biases in betting explains why it feels so convincing.

xG in betting models, and why arbitrage does not need it

In a forecasting model, xG is an input, not an answer. A typical approach turns a team's expected goals into a goal distribution with the Poisson formula, which goes back to work by Maher (1982) and Dixon and Coles (1997). If a model expects the home side to score 1.4 goals:

P(k goals) = e−λ × λk ÷ k!, with λ = 1.4
P(0) = 0.247 · P(1) = 0.345 · P(2) = 0.242

Combine the two teams' distributions and you get probabilities for 1X2, totals and correct scores. Those probabilities then have to beat the bookmaker's price after the margin, and over thousands of bets. That is the hard part of value betting, which is not yet in ArbVantage (coming soon).

Sure bets work differently. You do not forecast the match at all. When two bookmakers disagree enough on the price, you back every outcome so that, if all legs are accepted at the shown odds and stand, the payout is about the same whatever happens: a small margin, usually 1 to 2% of the total stake. No xG, no model, no opinion on who plays better. The risks are practical instead: odds change before the second leg, bets get voided and accounts get limited.

Frequently asked questions

What does an xG of 1.5 mean?

It means the shots a team took would, on average, produce 1.5 goals, based on similar shots in the past. It is not a forecast for the next match and not a probability of winning.

Can I beat the bookmakers with xG?

Not with public xG alone. It is free and widely read, so it is already part of the information in the odds. At best it helps you understand a price; any edge would need better data or a better model, tested over thousands of bets.

Why do xG values differ between websites?

Each data provider uses its own model with different inputs, such as body part, pass type or defender positions. Compare xG only from the same source.

Do I need xG for sure bets?

No. A sure bet comes from price differences between bookmakers, not from a forecast. Its risks are practical: odds changes, voided bets and account limits.

Sources

• Stats Perform: Expected goals (xG), definition and model
• The Analyst (Opta): Opta football stats definitions
• Maher, M. J. (1982): Modelling association football scores. Statistica Neerlandica 36(3)
• Dixon, M. J., Coles, S. G. (1997): Modelling association football scores and inefficiencies in the football betting market. JRSS C 46(2)
• Angelini, G., De Angelis, L. (2019): Efficiency of online football betting markets. International Journal of Forecasting 35(2)

As of October 2026. Calculations by ArbVantage; the match example is illustrative.

Betting involves risk, sure bets included: odds change, bets get voided and accounts get limited. Only bet money you can afford to lose. Our responsible gambling page lists free help.

Not betting, legal or tax advice. 18+. As of October 2026. See our disclaimer.